Re:How Figma's databases team lived to tell the scale
by esafak on Thursday March 14, 2024 @02:00PM
(#39707066)
This is such a familiar story. Company starts with a centralized database, runs into scaling problems, then spends man-years sharding it. Just use a distributed database.
[ Parent
| on HN ]
Re:How Figma's databases team lived to tell the scale
by dang on Thursday March 14, 2024 @02:19PM
(#39707249)
"Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something."
Re:How Figma's databases team lived to tell the scale
by jack_riminton on Thursday March 14, 2024 @02:05PM
(#39707106)
If every startup used every scale-proof method available from the beginning they'd never have the time or resources to build the products to get the customers that would require them to use the scale-proof methods
Re:How Figma's databases team lived to tell the scale
by esafak on Thursday March 14, 2024 @02:09PM
(#39707150)
I agree with your general point but there are relational distributed databases, open source and "serverless", that are compatible with MySQL (planetscale, tidb, vitess...) and postgres (citus, cockroachdb...)
edited for examples.
Re:How Figma's databases team lived to tell the scale
by mplanchard on Thursday March 14, 2024 @02:15PM
(#39707196)
And many of these are substantially more expensive than RDS, or have terrible INSERT performance, or require making the correct decisions really early about what to use as partition keys. There's no free lunch, sadly.
Re:How Figma's databases team lived to tell the scale
by noja on Thursday March 14, 2024 @02:11PM
(#39707169)
One reason they don't is unhelpful comments like these: alluding to products without naming them.
Re:How Figma's databases team lived to tell the scale
by ofrzeta on Thursday March 14, 2024 @02:18PM
(#39707238)
From the article: "During our evaluation, we explored CockroachDB, TiDB, Spanner, and Vitess". Three of them are open source if I am not mistaken.
Re:How Figma's databases team lived to tell the scale
by Closi on Thursday March 14, 2024 @02:22PM
(#39707294)
I don't think these were mature products when Figma started to be developed in 2012.
Re:How Figma's databases team lived to tell the scale
by Closi on Thursday March 14, 2024 @02:05PM
(#39707104)
This assumes that using a distributed database from the start doesn't offer it's own penalties/drawbacks (particularly in 2016).
Particularly considering as per the figma note, they consider their data highly relational (i.e. presumably this means lots of joins in queries)
What database would you have chosen?
Re:How Figma's databases team lived to tell the scale
by esafak on Thursday March 14, 2024 @02:07PM
(#39707132)
That's a fair response for then, but not today where you're spoiled for choice. Spanner was released as a service in 2017, which isn't far off.
Re:How Figma's databases team lived to tell the scale
by prisenco on Thursday March 14, 2024 @04:36PM
(#39708660)
Spanner is quite expensive though.
Re:How Figma's databases team lived to tell the scale
by Closi on Monday March 18, 2024 @04:45PM
(#39749788)
Well Spanner has it's own disadvantages, like having your entire stack tightly coupled to a google cloud product (i.e. you have to sacrifice being cloud agnostic).
Re:How Figma's databases team lived to tell the scale
by mamcx on Thursday March 14, 2024 @02:16PM
(#39707218)
There are many kinds of apps that use of "distributed" databases is an antipattern...and useless to make them scale.
MOST apps did not need more than "tenant-per-company" and then, maybe, distributed among a small set of servers.
Is it just that the kind of app that shows here (and has this problems) are just niche apps like social networks and complex chat apps.
Re:How Figma's databases team lived to tell the scale
by battwell on Thursday March 14, 2024 @02:20PM
(#39707264)
Doesn't seem a crazy way to start a company. RDS will have scaling problems but is very mature and fairly easy to use early on when you're working on your MVP
I've used CRDB early on at a startup. There was some overhead. You don't get all the nice PSQL features
Although it did save us a giant migration later on